On Mon, Apr 08, 2013 at 12:31:22AM +0200, Jean-S?bastien P?dron wrote: > On 07.04.2013 21:05, Konstantin Belousov wrote: > > I suspect that the issue is with the drm2/i915/intel_dp.c. > > You might try to use shift to 7 instead of 8 for address in the > > intel_dp_aux_native_read/write, but I am not sure. > > Based on your remark, I proposed the following patch to Olivier: > http://people.freebsd.org/~dumbbell/radeonkms/drm-i915-intel_iic.b.patch > > And it worked for him! > > If I understand correctly, this "aux channel" emulates a I2C bus. The > patch right-shifts back the slave address (in drm_dp_iic_helper.c). This > address is then used in intel_dp_i2c_aux_ch() (intel_dp.c). > > If you agree, I can commit it.
I see no reason to disagree. Please commit. As a side note, it is a miracle that DP connector KMS works at all. I had no DP hardware when I did the port, and Linux driver got a huge amount of bugfixes and churn in this area, which changes were only partially imported. I think that there is a complete rework of the modesetting sequence for Linux, either committed or going to be committed, motivated by the non-conformance of the DP and eDP handling.
pgp8xZmcJKOxF.pgp
Description: PGP signature
